Bird repelling method based on visual tracking

Abstract

The application relates to a bird driving method based on visual tracking, which solves the technical problems that in the prior art, management personnel cannot find birds in time sometimes, laser and sound emitted by a bird driving device cannot be accurately aimed at the birds, and on-site duty cost of the management personnel is high. Birds are recognized and tracked through a monocular camera, a firing angle of a gas gun is automatically adjusted, and the birds are dispersed by aiming at the birds.

Description

Technical Field

[0001] The present invention relates to the technical field of bird repelling, and in particular to a bird repelling method based on visual tracking. Background Art

[0002] As we all know, as the nation's ecological environment improves, the bird population is increasing. Birds not only severely impact crop yield and quality but can also cause power line failures. Bird strikes are also a significant safety issue for the aviation industry. To mitigate the harm posed by birds, various bird repellent devices are now in use, including laser repellers, sonic repellers, and gas cannons.

[0003] Currently, the way to trigger bird-repelling equipment is for management personnel to patrol locations such as airports, farmlands, orchards, and overhead power lines, and immediately activate the equipment to repel birds once they spot them. This method has the following technical issues:

[0004] (1) Managers sometimes fail to detect birds in time, or even fail to detect birds at all, leading to hazards.

[0005] (2) The laser and sound emitted by the bird-repelling equipment cannot accurately target birds.

[0006] (3) Management personnel are required to work on-site, resulting in high labor costs. Summary of the Invention

[0007] The present invention is designed to solve the technical problems of existing bird-repelling technology, such as managers sometimes cannot find birds in time, the lasers and sounds emitted by bird-repelling equipment cannot accurately target birds, managers need to be on duty on site, and labor costs are high. It provides a bird-repelling method based on visual tracking using visual recognition technology.

[0008] The present invention provides a bird-repelling method based on visual tracking, comprising the following steps:

[0009] The first step is that the bird enters the field of view of the monocular camera;

[0010] In the second step, a video stream is obtained through a monocular camera. The controller reads each frame of the video stream, detects the bird using the frame difference method, and outputs the pixel points representing the bird. The process of detecting the bird using the frame difference method generates a maximum bounding rectangle for selecting the bird. The image information of this rectangle is input into the KCF tracker to continuously obtain the maximum bounding rectangle.

[0011] In the camera coordinate system, the vertical distance Z between the bird and the camera in the field of view plane of the monocular camera is calculated according to the principle of similar triangles using the following formula (1):

[0012] (1);

[0013] In formula (1), f is the focal length of the monocular camera; x is the size of the bird image; K is the actual area size of the bird;

[0014] Calculated by the following formula (3) :

[0015] (3);

[0016] In formula (3), D 12 It is through the formula Calculated, d 12 Indicates the horizontal pixel distance between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangular frame;

[0017] Calculated by the following formula (4) :

[0018] (4);

[0019] In formula (4), D 21 It is through the formula Calculated, d 21 Indicates the vertical distance in pixels between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangle;

[0020] Next, the actual distance d between the bird and the monocular camera is calculated using the following formula (5):

[0021] d= (5);

[0022] In the third step, the controller controls the pan-tilt motion through visual servoing, and the pan-tilt rotates in the horizontal direction. , the gimbal rotates in the pitch direction The pan-tilt action drives the monocular camera to rotate, so that the bird is always within the monitoring range of the monocular camera, allowing the monocular camera to continuously track the bird.

[0023] Step 4: When the actual distance d is less than the set trigger threshold, the current gimbal horizontal rotation angle is obtained. and the current gimbal vertical rotation angle , the controller controls the rotation device to adjust the firing angle of the gas gun, so that the rotation device rotates to θw' in the horizontal direction, θw'= + , in the pitch direction, turn to θh', θh'= + , then start the gas to disperse the birds.

[0024] Preferably, the bird-repelling operation ends when the bird is no longer within the field of view of the monocular camera.

[0025] The present invention also provides a bird-repelling method based on visual tracking, comprising the following steps:

[0026] The first step is that the bird enters the field of view of the monocular camera;

[0027] In the second step, a video stream is obtained through a monocular camera. The controller reads each frame of the video stream, detects the bird using the frame difference method, and outputs the pixel points representing the bird. The process of detecting the bird using the frame difference method generates a maximum bounding rectangle for selecting the bird. The image information of this rectangle is input into the KCF tracker to continuously obtain the maximum bounding rectangle.

[0028] In the camera coordinate system, the vertical distance Z between the bird and the camera in the field of view of the monocular camera is calculated according to the principle of similar triangles using the following formula (2):

[0029] (2);

[0030] In formula (2), f is the focal length of the monocular camera, K represents the actual area of ​​the bird; s represents the area of ​​the bird's image;

[0031] Calculated by the following formula (3) :

[0032] (3);

[0033] In formula (3), D 12 It is through the formula Calculated, d 12 Indicates the horizontal pixel distance between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangular frame;

[0034] Calculated by the following formula (4) :

[0035] (4);

[0036] In formula (4), D 21 It is through the formula Calculated, d 21 Indicates the vertical distance in pixels between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangle;

[0037] Next, the actual distance d between the bird and the monocular camera is calculated using the following formula (5):

[0038] d= (5);

[0039] In the third step, the controller controls the pan-tilt motion through visual servoing, and the pan-tilt rotates in the horizontal direction. , the gimbal rotates in the pitch direction The pan-tilt action drives the monocular camera to rotate, so that the bird is always within the monitoring range of the monocular camera, allowing the monocular camera to continuously track the bird.

[0040] Step 4: When the actual distance d is less than the set trigger threshold, the current gimbal horizontal rotation angle is obtained. and the current gimbal vertical rotation angle , the controller controls the rotation device to adjust the firing angle of the gas gun, so that the rotation device rotates to θw' in the horizontal direction, θw'= + , in the pitch direction, turn to θh', θh'= + , then start the gas to disperse the birds.

[0041] Preferably, when the bird is no longer within the field of view of the monocular camera, the bird-repelling operation is terminated, and the accumulated data is trained to achieve the prediction of the bird's habitat.

[0042] Preferably, the process of training the accumulated data and predicting the bird's habitat is:

[0043] Step 1, data accumulation of feature 1:

[0044] When the bird is detected for the first time, data accumulation begins. By obtaining the rectangular frame, the corresponding Secondly, the current horizontal rotation angle of the gimbal can be obtained through the motor encoder in the gimbal ,because is the horizontal angle of the bird away from the center of the field of view, so the actual horizontal angle of the bird θw'= + , with the fort as the center, divide 360° into 16 directions, each direction contains 22.5°, direction 1 contains 0-22.5°, direction 2 contains 22.5°-45°, and so on. Then calculate the corresponding direction of the bird's actual horizontal angle θw' according to the divided area. Use sixteen variables a1, a2, a3…a16 to record the number of times the bird appears in each direction. When the actual horizontal angle θw' of the bird is within the range of one of the directions, the counting condition is met. The controller receives a trigger pulse and increases the variable corresponding to the direction, completing the accumulation of bird data.

[0045] Repeat the above process and record the direction and the corresponding number of times the bird appears:

[0046] Step 2, data accumulation of feature 2:

[0047] After the bird is dispersed, the bird is tracked continuously. When the bird exceeds the maximum detection range, data accumulation begins. The corresponding rectangular frame at that moment is calculated. , and obtain the horizontal rotation angle of the pan-tilt through the motor encoder in the pan-tilt , calculating the bird's actual horizontal angle θw' at that time. Then, in the same way, record the direction the bird left from and the corresponding number of times; Feature 1 will predict the direction where the bird appeared most frequently as the habitat, and Feature 2 will also predict the direction where the bird left most frequently as the habitat. The Bayesian algorithm determines whether to use Feature 1 or Feature 2 for prediction.

[0048] The invention has the following advantages: it uses visual recognition technology to automatically identify and track birds, accurately targeting and dispersing them, significantly improving the effectiveness and efficiency of bird repelling. It eliminates the need for managers to detect birds with the naked eye and eliminates the need for managers to be on duty, thus reducing labor costs.

[0049] To prevent birds from flying in advance, such as before an airplane takes off, it is necessary to fire a cannon in advance to prevent bird interference. Therefore, when birds are not present, a gas cannon is fired in the direction of the bird habitat to prevent birds from flying. To prevent birds from flying in advance, the direction of the bird habitat is predicted by accumulating data based on two features: (1) the direction and number of birds when they enter the detection range; (2) after the gas cannon is fired, the dispersed birds are continuously tracked and the last direction and number of birds when they leave the detection range are accumulated.

[0050] Further features and aspects of the present invention will be clearly described in the following description of specific embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a structural diagram of a bird repellent device;

[0052] Figure 2 This is a block diagram of the visual recognition system and the control of the gas cannon;

[0053] Figure 3 It is a flow chart of the bird-repelling method of the present invention;

[0054] Figure 4 It is a schematic diagram of the bird's position in the field of view plane of the monocular camera;

[0055] Figure 5 This is a schematic diagram of the bird's image within the field of view of a monocular camera;

[0056] Figure 6This is the principle diagram for calculating the actual distance between the bird and the monocular camera;

[0057] Figure 7 This is a simple bird-repelling top-down diagram. The center of the circle is the position of the turret, the radius is the distance that the monocular camera can observe, the field of view is divided into 16 parts, and the triangulation position detects the area where the bird appears;

[0058] Figure 8 This is a simple bird-chasing diagram from above. The center of the circle is the position of the turret, the radius is the distance that the monocular camera can observe, and the field of view is divided into 16 parts. The circle is the area where the bird is located after it is chased away and tracked.

[0059] Figure 9 It is a flowchart of the Bayesian algorithm.

[0060] Explanation of symbols in the figure:

[0061] 1. Base, 2. Pan / tilt, 3. Rotating device, 4. Monocular camera, 5. Gas cannon, 6. Controller. DETAILED DESCRIPTION

[0062] The present invention will be further described in detail below with reference to the accompanying drawings using specific embodiments.

[0063] like Figure 1 and 2 As shown, the bird repellent device includes a base 1, a pan-tilt head 2, a rotating device 3, a monocular camera 4, a gas cannon 5, and a controller 6. The pan-tilt head 2 and the rotating device 3 are respectively mounted on the base 1, the monocular camera 4 is mounted on the pan-tilt head 2, and the gas cannon 5 is connected to the rotating device 3. The monocular camera 4 is connected to the controller for communication, the pan-tilt head 2 is electrically connected to the controller 6, the rotating device 3 is electrically connected to the controller 6, and the gas cannon 5 is electrically connected to the controller 6.

[0064] The controller 6 can control the movement of the rotating device 3 , and the movement of the rotating device 3 can drive the gas gun 5 to rotate so as to adjust the firing angle.

[0065] The controller 6 can control the movement of the pan-tilt head 2, which can drive the monocular camera 4 to rotate and adjust the shooting range, so that the camera 4 tracks the bird.

[0066] The gas gun 5 is a prior art, and its specific structure, working principle and process are not described in detail here. For details, please refer to the utility model patent with authorization announcement number CN 210226699 U.

[0067] refer to Figure 3 The bird-repelling method disclosed in the present invention mainly comprises the following steps:

[0068] In the first step, the bird enters the field of view of the monocular camera 4.

[0069] In the second step, the monocular camera 4 acquires a video stream. Controller 6 reads each frame of the video stream, detects the bird using a frame difference method, and outputs the pixel points representing the bird. This process generates a maximum bounding rectangle for the bird. This rectangle's image information is fed into the KCF tracker, which continuously obtains the maximum bounding rectangle.

[0070] The full name of KCF is Kernel Correlation Filter kernel correlation filtering algorithm.

[0071] like Figure 4 As shown, in the camera coordinate system, according to the principle of similar triangles, the following equation is obtained:

[0072] ;

[0073] is the focal length f of the monocular camera 4; is the vertical distance Z between the bird and the origin O, that is, the vertical distance between the bird and the camera in the field of view plane of the monocular camera; is the size x of the bird image; is the actual size of the bird K; therefore, the vertical distance Z between the bird and the camera in the field of view of the monocular camera can be calculated by the following formula (1);

[0074] (1)

[0075] From formula (1), we can see that

[0076] ;

[0077] The length of the bird image is the product of the number of pixels along the Y-axis and the length of a single pixel along the Y-axis.

[0078] In order to more accurately calculate the vertical distance Z between the bird and the camera in the field of view of the monocular camera, Z is calculated using the following formula (2):

[0079] (2);

[0080] In formula (2), K represents the actual area of ​​birds that often appear in application scenarios (such as airports); s represents the area of ​​the bird image, which is the product of the total number of pixels occupied by the bird image and the area of ​​a single pixel.

[0081] refer to Figure 5 、 6 , d 12 The horizontal distance in pixels between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangular frame (number of horizontal pixels × horizontal length of a single pixel), d21 Indicates the vertical pixel distance between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangle (number of vertical pixels × vertical length of a single pixel).

[0082] By formula Calculate D 12 , D 12 Indicates the actual horizontal distance between the center of the monocular camera's field of view and the center of the bird's maximum circumscribed rectangular frame.

[0083] By formula Calculate that D 21 , D 21 Indicates the actual vertical distance between the center of the monocular camera's field of view and the center of the bird's maximum circumscribed rectangular frame.

[0084] Calculated by the following formula (3) :

[0085] (3);

[0086] Calculated by the following formula (4) :

[0087] (4);

[0088] Next, the actual distance d between the bird and the monocular camera 4 is calculated using the following formula (5):

[0089] d= (5)

[0090] In the third step, the controller 6 controls the movement of the pan / tilt head 2 through visual servoing. As the horizontal angle of PTZ 2, As the pitch angle of the gimbal, gimbal 2 rotates in the horizontal direction , gimbal 2 rotates in the pitch direction The movement of the pan-tilt head 2 drives the monocular camera 4 to rotate, and the bird is always within the monitoring range of the monocular camera (keeping the center of the bird's largest circumscribed rectangular frame coincident with the center of the monocular camera's field of view), so that the monocular camera 4 can continuously track the bird.

[0091] In the fourth step, when the bird enters the range of 40m, that is, when the actual distance d is less than the set trigger threshold 40m, the current horizontal rotation angle θw and the current vertical rotation angle θh of the pan-tilt 2 are obtained through the motor encoder of the pan-tilt 2. The controller 6 controls the rotation device 3 to adjust the firing angle of the gas cannon 5, so that the rotation device 3 rotates horizontally to θw', θw'= + , in the pitch direction, turn to θh', θh'= + , then start the gas cannon 5, aim at the bird and fire the gas cannon to disperse the bird.

[0092] When the bird is not within the field of view of the monocular camera 4, the frame difference method will not generate a bounding rectangle when detecting the bird, and the bird repelling operation is terminated. The data accumulated during the entire process can be used for training to predict the bird's habitat.

[0093] The following describes a specific method for predicting bird habitats using accumulated data:

[0094] (1) Data accumulation of feature 1

[0095] refer to Figure 7 When the bird is detected for the first time, data accumulation begins. By obtaining the rectangular frame, the corresponding Secondly, the current horizontal rotation angle of the gimbal can be obtained through the motor encoder in gimbal 2 ,because is the horizontal angle of the bird from the center of the field of view, so the actual horizontal angle of the bird is

[0096] θw'= + With the fort as the center, 360° is divided into 16 azimuths, each encompassing 22.5°. Azimuth 1 covers 0-22.5°, Azimuth 2 covers 22.5-45°, and so on. The bird's actual horizontal angle θw' is then calculated based on the divided areas to determine the corresponding azimuth. Sixteen variables, a1, a2, a3…a16, are used to record the number of times the bird appeared in each azimuth. When the bird's actual horizontal angle θw' falls within the range of any of these azimuths, the counting condition is met. The controller receives a trigger pulse and increments the variable corresponding to that azimuth, completing the bird data accumulation.

[0097] Repeat the above process and record the direction where the bird appears and the corresponding number of times.

[0098] (2) Data accumulation of feature 2:

[0099] refer to Figure 8 After the bird is dispersed, the bird is tracked continuously. When the bird exceeds the maximum detection range, data accumulation begins. The corresponding rectangular frame at that moment is calculated. , and obtain the horizontal rotation angle of the pan-tilt through the motor encoder in the pan-tilt , calculate the actual horizontal angle θw' of the bird at that time. Then record the direction and corresponding number of times the bird left in the same way.

[0100] Feature 1 will predict the direction where the bird appears most frequently as the habitat, for example Figure 7 In the direction 7, feature 2 will also predict that the direction where the bird leaves the most times is the habitat, for example Figure 8 Position 1 in.

[0101] The Bayesian algorithm calculates the probability that feature 1 or feature 2 is correct. In other words, it determines which feature is most accurate and uses that feature. The idea is to continuously fire at feature 1 or feature 2. If feature 1 performs well, the weight of feature 1 is increased; if feature 2 performs well, the weight of feature 2 is increased.

[0102] To measure the quality of our predictions, we set a time threshold: the average time between bird appearances, provided the birds are not driven away. If, after firing at the predicted location, the bird does not appear within the time threshold, we consider the predicted habitat a success.

[0103] After accumulating data, the data is iteratively updated and weighted through machine learning. By selecting a suitable probability model based on the accumulated data, the bird's habitat can be predicted to prevent bird repelling.

[0104] The Bayesian algorithm is used to predict which direction is most likely to be the habitat. Naive Bayes is a machine learning algorithm based on probability statistics, which is used for multi-classification. When faced with independent features, the first step is to determine the feature attributes and train the data set. For a given x The posterior probability distribution can be calculated through the learned model In the formula, because the denominator is equivalent to the probability of X existing in the database, P(X) is a constant for any item to be classified; then calculate the posterior probability When we consider only the numerator, we only need to consider the molecule. The main process is:

[0105] Step (1) First accumulate data

[0106] Accumulation process: Pre-fire at the locations predicted by features 1 and 2, and record the prediction results. If the prediction is for feature 1, then feature 1 is accumulated as correct, while feature 2 is accumulated as incorrect. If the prediction is successful (the bird does not arrive within the threshold time), then the habitat is accumulated as yes, and if the prediction fails, then the habitat is accumulated as no.

[0107] For example, if feature 1 predicts direction 7 and feature 2 predicts direction 1, the data accumulation process is shown in the following table:

[0108] Feature 1 Feature 2 Habitat right wrong yes right wrong no wrong right yes wrong right no

[0109] The first two examples in the table represent premature firing at direction 7, as predicted by feature 1. Therefore, feature 1 is correct and feature 2 is incorrect. The first example in the table represents a successful prediction after firing (the bird did not arrive within the threshold time), so the habitat is correct. The second example represents a failed prediction after firing (the bird arrived within the threshold time). Similarly, the last two examples in the table represent firing at direction 1, as predicted by feature 2. The third example in the table represents a successful prediction, while the fourth example represents a failed prediction. This continues as data is accumulated.

[0110] In step (2), the probability that the habitat predicted by feature 1 is correct can be calculated through the accumulated data, that is, when feature 1 is correct and feature 2 is wrong, the probability that the habitat is correct can be calculated.

[0111] Feature 1 Feature 2 Habitat right wrong yes

[0112] There are usually flocks of birds in bird habitats. Firing cannons at bird habitats before the plane takes off can disperse the flocks and prevent interference from the flocks.

[0113] It should be noted that the present invention is described using a gas cannon as an example. Those skilled in the art will appreciate that the present invention can also be applied to laser bird repellers, sonic bird repellers, and the like.

[0114] Bird activities can cause overhead power transmission line failures. Birds moving around overhead power transmission line towers, their activity tracks, and bird defecation can easily cause power transmission line tower failures. The present invention can also be used to repel birds from overhead power transmission lines.

[0115] The above description is only about the preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations.

Claims

1. A bird-repelling method based on visual tracking, characterized in that: The following steps are involved: The first step is that the bird enters the field of view of the monocular camera; In the second step, a video stream is obtained through a monocular camera. The controller reads each frame of the video stream, detects the bird using the frame difference method, and outputs the pixel points representing the bird. The process of detecting the bird using the frame difference method generates a maximum bounding rectangle for selecting the bird. The image information of this rectangle is input into the KCF tracker to continuously obtain the maximum bounding rectangle. In the camera coordinate system, the vertical distance Z between the bird and the camera in the field of view plane of the monocular camera is calculated according to the principle of similar triangles using the following formula (1): （1）； In formula (1), f is the focal length of the monocular camera; x is the size of the bird image; K is the actual area size of the bird; Calculated by the following formula (3) : （3）； In formula (3), D 12 It is through the formula Calculated; D 12 Indicates the horizontal pixel distance between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangular frame; Calculated by the following formula (4) : （4）； In formula (4), D 21 It is through the formula Calculated; d 21 Indicates the vertical distance in pixels between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangle; Next, the actual distance d between the bird and the monocular camera is calculated using the following formula (5): d= （5）； In the third step, the controller controls the pan-tilt motion through visual servoing, and the pan-tilt rotates in the horizontal direction. , the gimbal rotates in the pitch direction The pan-tilt action drives the monocular camera to rotate, so that the bird is always within the monitoring range of the monocular camera, allowing the monocular camera to continuously track the bird. Step 4: When the actual distance d is less than the set trigger threshold, the current gimbal horizontal rotation angle is obtained. and the current gimbal vertical rotation angle , the controller controls the rotation device to adjust the firing angle of the gas gun, so that the rotation device rotates to θw' in the horizontal direction, θw'= + , in the pitch direction, turn to θh', θh'= + , then start the gas to disperse the birds.

2. The bird-repelling method based on visual tracking according to claim 1, characterized in that: When the bird is out of the field of view of the monocular camera, the bird-repelling operation ends.

3. A bird-repelling method based on visual tracking, characterized in that: The following steps are involved: The first step is that the bird enters the field of view of the monocular camera; In the second step, a video stream is obtained through a monocular camera. The controller reads each frame of the video stream, detects the bird using the frame difference method, and outputs the pixel points representing the bird. The process of detecting the bird using the frame difference method generates a maximum bounding rectangle for selecting the bird. The image information of this rectangle is input into the KCF tracker to continuously obtain the maximum bounding rectangle. In the camera coordinate system, the vertical distance Z between the bird and the camera in the field of view plane of the monocular camera is calculated according to the principle of similar triangles using the following formula (2): （2）； In formula (2), f is the focal length of the monocular camera, K represents the actual area of ​​the bird; s represents the area of ​​the bird's image; Calculated by the following formula (3) : （3）； In formula (3), D 12 By formula Calculated, d 12 Indicates the horizontal pixel distance between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangular frame; Calculated by the following formula (4) : （4）； In formula (4), D 21 By formula Calculated, d 21 Indicates the vertical distance in pixels between the center of the monocular camera's field of view and the center of the bird's largest circumscribed rectangle; Next, the actual distance d between the bird and the monocular camera is calculated using the following formula (5): d= （5）； In the third step, the controller controls the pan-tilt motion through visual servoing, and the pan-tilt rotates in the horizontal direction. , the gimbal rotates in the pitch direction The pan-tilt action drives the monocular camera to rotate, so that the bird is always within the monitoring range of the monocular camera, allowing the monocular camera to continuously track the bird. Step 4: When the actual distance d is less than the set trigger threshold, the current gimbal horizontal rotation angle is obtained. and the current gimbal vertical rotation angle , the controller controls the rotation device to adjust the firing angle of the gas gun, so that the rotation device rotates to θw' in the horizontal direction, θw'= + , in the pitch direction, turn to θh', θh'= + , then start the gas to disperse the birds.

4. The bird-repelling method based on visual tracking according to claim 3, characterized in that: When the bird is no longer within the field of view of the monocular camera, the bird-scaring operation ends and the accumulated data is trained to predict the bird's habitat.

5. The bird-repelling method based on visual tracking according to claim 4, characterized in that: The process of training the accumulated data and predicting the bird's habitat is as follows: Step 1, data accumulation of feature 1: When the bird is detected for the first time, data accumulation begins. By obtaining the rectangular frame, the corresponding Secondly, the current horizontal rotation angle of the gimbal can be obtained through the motor encoder in the gimbal ,because is the horizontal angle of the bird away from the center of the field of view, so the actual horizontal angle of the bird θw'= + , with the fort as the center, divide 360° into 16 directions, each direction contains 22.5°, direction 1 contains 0-22.5°, direction 2 contains 22.5°-45°, and so on; then calculate the corresponding direction of the bird's actual horizontal angle θw' according to the divided areas; Sixteen variables, a1, a2, a3…a16, are used to record the number of times a bird appears in each direction. When the actual horizontal angle θw' of the bird is within the range of one of the directions, the counting condition is met. The controller receives a trigger pulse and increases the variable corresponding to the direction, completing the accumulation of bird data. Repeat the above process and record the direction and the corresponding number of times the bird appears; Step 2, data accumulation of feature 2: After the bird is dispersed, the bird is tracked continuously. When the bird exceeds the maximum detection range, data accumulation begins. The corresponding rectangular frame at that moment is calculated. , and obtain the horizontal rotation angle of the pan-tilt through the motor encoder in the pan-tilt , calculate the actual horizontal angle θw' of the bird at that time, and then record the direction and corresponding number of times when the bird left in the same way; Feature 1 will predict the direction where the bird appears most frequently as its habitat, and feature 2 will also predict the direction where the bird leaves most frequently as its habitat. The Bayesian algorithm is used to determine whether to make a prediction based on feature 1 or feature 2.

Images